2021 Korea Astronomy Machine Learning

Tuesday 6 Apr 2021, 13:30 → 17:00 Asia/Seoul

University of Seoul, NSRI(Natural Science Research Institute)

Inkyu Park (University of Seoul)

Dear all,

We are pleased to announce the next meeting in the Korea Astronomy Machine Learning (KAML) series. It will take place on 6th April, hosted by the University of Seoul (서울시립대학교). This meeting is part of a seres where astrophysics researchers, who are keen to make use of machine learning in their research, can meet to discuss, learn new techniques, and create new projects.

These meetings are open to anyone, even those who have not previously attended such a meeting. The next meeting will introduce the KAML meetings, and include new opportunities for projects or collaborations. It will also include two review talks by Dr Min-Su Shin and Dr Ryan Keeley (KASI). These reviews will be aimed at the machine learning novice, who is keen to learn more.

The meeting will be hosted by the University of Seoul, but participants can also join online using Zoom.

https://uos-ac-kr.zoom.us/j/8264902652

1 Welcome Remark (David Parkinson & Inkyu Park)

2 Applications of Machine Learning Algorithms and their Challenges in Astronomy

Astronomy has a long history of exploiting modern hardware architectures and algorithms in its data analysis research. The most recent trend includes applications of cutting-edge machine learning algorithms. In this talk, I will introduce several application cases of machine learning algorithms, focusing on data science with big survey data, and some challenges and problems in these applications will be presented with some possible ways to tackle these issues.

Speaker: Minsu Shin

3 Gaussian Process Regression in Cosmology

Gaussian process regression is a data driven statistical tool that is useful for both reconstructions and model selection. Especially in this era of precision cosmology where tensions abound, this example of model-independent statistical methods can be crucial for identifying the new physics or systematics that is the explanation for the tensions. In this talk I will review the formalism of Gaussian process regression. In particular, I will discuss the “dos and donts” of this statistical tool and point out common pitfalls and how to avoid them. Further, I will discuss recent interesting cosmological results found using Gaussian process regression.

Speaker: Ryan Keeley

4 Break

5 Discussion #1 : Summary of previous KAML projects

Machine Learning - Zoom - Apr 2021.pdf

6 Discussion #2 : New KAML collaborational projects

autoencoder_sketch.pdf

axion 21cm

SKA Science Data Challenge.pdf